Daniel Graham: The Efficient Artist: Statistical Regularities in Art and Their Relationships with Visual Coding
Talk by Daniel Graham, Post-Doctoral Researcher, Dept. of Mathematics, Dartmouth College, Hanover, NH. Given to the UC Berkeley Vision Science Oxyopia seminar on November 20, 2009 at UC Berkeley. Host is Bruno Olshausen.
Producer UC Berkeley Vision ScienceAudio/Visual sound, color
This talk describes statistical transformations artists perform in order to depict the natural visual world. In contrast to natural scenes, artworks are a vastly reduced stimulus in a statistical sense. But art can yet be perceived as representing a specific scene or category of scene. The goal of this work is to quantify artists’ shared strategies of representing the world. I will describe regularities in luminance and spatial statistics for art and show evidence of links between these regularities and behavioral measures of perception. My work suggests artists may transform natural scene statistics so as to efficiently stimulate the human visual system and work is now underway to isolate those features of artistic representations that could map onto human visual coding strategies. Since only humans make art, and since this ability appears related to our unique perceptual processing, rather than to unique motor skills, this approach provides new tools for understanding human visual representations.